Social media site Instagram is bringing in two new anti-bullying features after being criticised for not doing enough.
It has started rolling out artificial intelligence, which flags up potentially offensive material before someone posts it.
It allows users to 'undo' potentially harmful comments, and will not send a notification to the recipient.
In a statement, head of Instagram Adam Mosseri says: "In the last few days, we started rolling out a new feature powered by AI that notifies people when their comment may be considered offensive before it's posted.
"This intervention gives people a chance to reflect and undo their comment and prevents the recipient from receiving the harmful comment notification.
"From early tests of this feature, we have found that it encourages some people to undo their comment and share something less hurtful once they have had a chance to reflect."
The company is also working on a service called 'Restrict', which can protect users accounts from unwanted interactions with someone else - without telling them.
Mr Mosseri explains: "We've heard from young people in our community that they're reluctant to block, unfollow, or report their bully because it could escalate the situation, especially if they interact with their bully in real life.
"Some of these actions also make it difficult for a target to keep track of their bully's behavior.
"We wanted to create a feature that allows people to control their Instagram experience, without notifying someone who may be targeting them.
"Soon, we will begin testing a new way to protect your account from unwanted interactions called 'Restrict'.
"Once you Restrict someone, comments on your posts from that person will only be visible to that person."
Users can also choose to make a restricted person's comments visible to others by approving their comments.
But 'Restricted' users will not be able to see when another person is active on Instagram, or when their direct messages have been read.